it' s all about the old idea of taking a sound input and producing images from it. our approach is mostly interested in finding algorithms that perform a transformation of the original data into a different representation space. the hope is to find such spaces together with appropriate basis functions that will result in representations that are aesthetically meaningful and interesting.
0:02:10-0:02:15 0:11:40-0:11:52 0:21:45-0:21:55 0:26:20-0:26:39
samples taken from a concert by marcus schmickler at batofar, paris may 23,1999.
click on the above images to load the corresponding quicktime movie sample.
the above movie samples were obtained using a program that performs a one-dimensional windowed fft on the sampled sound input. 16 frequency-bands with a range of one byte are computed. these values are the input to a variety of algorithms, which range from graphing routines to parametric animations of sampled images. we have also used hidden markov models that were trained offline on the sound input and whose hideen states were used to generate images during the performance.
contact: constantin rothkopf.
last updated nov. 23, 2001